The Real Distinction Between Descriptions and Indexicals*

teorema Vol. XXIV/3, 2005, pp. 49-74

The Real Distinction between Descriptions and Indexicals*

Manuel García-Carpintero

ABSTRACT Some contemporary semantic views defend an asymmetry thesis concerning definite descriptions and indexicals. Semantically, indexicals are devices of singular reference; they contribute objects to the contents of the speech acts made with utterances including them. Definite descriptions, on the other hand, are generalized quantifiers, behaving roughly the way Russell envisaged in “On Denoting”. The asymmetry thesis depends on the existence of a sufficiently clear-cut distinction between semantics and pragmatics, because indexicals and descriptions are often used in ways that apparently contradict the asymmetry thesis; the semantics/pragmatics distinction is invoked to see behind the appearances. The paper critically examines arguments by Schiffer against the asymmetry thesis, based on referential uses of incomplete descriptions.

RESUMEN Algunas teorías semánticas contemporáneas defienden una tesis de asimetría respecto de las descripciones definidas y los indéxicos. Semánticamente, los indéxicos son instrumentos de referencia singular; contribuyen con objetos a los contenidos de los actos del habla que se llevan a cabo por medio de ellos. Las descripciones definidas, por su parte, son cuantificadores generalizados, que se comportan a grandes ras- gos como Russell indicara en “On Denoting”. La tesis de la asimetría depende de la existencia de una distinción suficientemente clara entre semántica y pragmática, dado que tanto los indéxicos como las descripciones definidas se usan a menudo de mane- ras que parecen contradecirla; se invoca entonces la distinción entre semántica y pragmática para reconciliarla con las apariencias. El trabajo examina críticamente ar- gumentos recientes de S. Schiffer contra la tesis de asimetría, que se apoyan en usos referenciales de descripciones incompletas.

In contemporary semantics an asymmetry thesis is hold concerning the behaviour of, respectively, definite descriptions and indexicals. Semantically, indexicals are devices of singular reference; they contribute objects to the contents of the speech acts made with utterances including them. Definite descriptions, on the other hand, are generalized quantifiers, behaving roughly the way Russell envisaged in “On Denoting”. The asymmetry thesis depends on the existence of a sufficiently clear- cut distinction between semantics and pragmatics, because indexicals and de-

49 50 Manuel García-Carpintero scriptions are often used in ways that apparently contradict the asymmetry thesis; the semantics/pragmatics distinction is invoked to see behind the appearances. There are Donnellan’s “referential uses” of descriptions [Donnellan (1966)]; and there are descriptive uses of indexicals, Kaplan’s “monsters” [Kaplan (1989a)]. Defenders of the asymmetry thesis hold that these are non- semantic phenomena, a form of conversational implicature; for descriptions, Grice [Grice (1969)], Kripke [Kripke (1977)], Evans [Evans (1982)] and Neale [Neale (1990)] provided supporting arguments. It is therefore no surprise that writers who question the relevance or the very existence of the semantics/pragmatics distinction (whom I refer to as pragmaticists) argue against the asymmetry thesis. In this paper I will critically examine arguments by one of these writers, Schiffer [Schiffer (1995)], based on referential uses of incomplete descriptions. In so far as his arguments are symptomatic of the pragmaticist (mis)conception of language, we can learn a more general lesson from its weaknesses.

I. THE ASYMMETRY THESIS STATED

There is a very simple way of making the semantics/pragmatics distinction and then stating relative to it the asymmetry thesis, which is unfortunately incorrect. According to this view, semantics has just to do with compositionally ascribing truth-conditions to sentences; anything involving speakers’ intentions is a matter for pragmatics to care about. From this perspective, indexicals are semantically like Millian names, characterized by Kripke thus: “According to Mill, a proper name is, so to speak, simply a name. It simply refers to its bearer, and has no other linguistic function. In particular, unlike a definite description, a name does not describe its bearer as possessing any special identifying properties” [Kripke (1979), pp. 239-40]. Larson and Segal [Larson and Segal (1996), pp. 197-221] extend this view to indexicals. Unfortunately, I do not think that the contrast that the asymmetry thesis claims to exist is that simple. I do not think that proper names are Mil- lian; I believe that it is part of their function to describe their referent as having special identifying properties, and that this is a semantic, linguistic matter. Intuitively, the same applies even more obviously to indexicals. Here is a reason. Competent speakers of English recognize the validity of the following inferences; if someone does not recognize their validity, that prima facie disqualifies him as a competent speaker of English:

(1) He is hungry

∴ Some male is hungry

The Real Distinction between Descriptions and Indexicals 51

(2) That pot is empty

∴ Some pot is empty

(3) You are angry

∴ Someone in the audience is angry

Competent speakers possessing the relevant concepts also recognize the tautological character of the following sentences:

(4) He is whoever is most salient male when that very token of ‘he’ is uttered

(5) That pot is whatever is most salient pot when that very token of ‘that pot’ is uttered

(6) You are whoever is addressed when that very token of ‘you’ is uttered

In order to capture this point (the point, as he put it, that there is a logic of indexicals) while still holding the asymmetry thesis, Kaplan [Kaplan (1989a), (1989b)] ascribed to indexicals not only a referent as content, but also descriptive features as character, and suggested a more complex way of making the semantics/pragmatics distinction than the one just outlined. Kaplan [Kaplan (1989b)] distinguished the descriptive features associated with indexicals constituting their character, which belong in a semantic account of a language, from other descriptive features associated with them (and from every descriptive feature associated with proper names), which in his view belong only in a metasemantic account. Stalnaker [Stalnaker (1997)] makes a similar distinction with a different terminology, between descriptive and foundational linguistic matters. I will borrow from Kaplan’s and Stalnaker’s suggestions. Lewis [Lewis (1983)] distinguished languages (and theoretical accounts thereof) understood as abstract objects, which may or may not be actually used, from languages seen as actually spoken by a given individual or population. In this vein I distinguish, firstly, a (linguistic-descriptive) theoretical characterization of an abstract language intended to have the constitutive features necessary and sufficient for it to count as a given natural language, i.e., to fully character- ize the language used by a given population; and, secondly, a (linguistic- foundational) theoretical characterization of facts such that a given abstract language is indeed used by a given population. Different views on what Schiffer [Schiffer (1993)] calls “the actual-language relation” would provide different elaborations on this distinction, but we can suggest some relatively non-contentious illustration. Thus, general psychological facts regarding memory, attention, and perceptual or procedural capacities will have to be 52 Manuel García-Carpintero invoked in a foundational characterization that a given population uses a given abstract language; for, together with language mastery, those general psychological capacities are a crucial causal factor of language use. Simi- larly, something like the Gricean conversational maxims will have to be invoked in a foundational characterization that a given population uses a given abstract language; for once again, together with language mastery they are, at least in some cases (those involving undisputed cases of conversational implicatures), a crucial causal factor of language use. None of them, however, belong in a descriptive account.1 This framework, I submit, provides a taxonomically (and therefore theoretically) more accurate way of tracing the semantics/pragmatics distinction than the simplistic one previously outlined: semantics is part of descriptive linguistics; pragmatics rather belongs in foundational linguistics. That this is more accurate can be seen in that, for instance, it allows us to plausibly contend both that moods (interrogative, imperative, indicative and so on) do not make truth-conditional contributions, and also that it is part of semantics, not pragmatics, to account for them. Davidsonian partisans of the view that semantics has just to do with the compositional determination of truth- conditions dispute this; but their view goes against intuition, Davidson’s [Davidson (1979)] arguments for it are fallacious, and in any case it should not be established just by a not merely stipulative definition.2 In addition to a non-truth-conditional view of the illocutionary forces signified by moods, the present proposal allows us to count conventional implicatures and presuppositions conventionally conveyed by expressions as phenomena for a semantic account to care about, not for a pragmatic one; and this is crucial for my purposes, because in my own view those phenomena provide the proper way to look at the linguistic descriptive meanings of referential expressions like indexicals (and proper names, but I will put them aside here). I turn now to outline the view, which I have developed in more detail elsewhere, and to state relative to it the asymmetry thesis in the form that I want to defend it [García-Carpintero (1998) and (2000)]. Mine is one among recent multi-propositional proposals contending that more than one proposition is usually expressed by an utterance. Dever [Dever (2001)] argues for this view concerning non-restrictive relative clauses, as in (7), and complex demonstratives, as in (8):

(7) Joan, who is mayor of Barcelona, might study Dennett’s philosophy

(8) That mayor of Barcelona might study Dennett’s philosophy

According to Dever’s proposal, an utterance of (7) is an assertion of a proposition to which the material in the appositive clause does not contribute; in addition, a second proposition is expressed in uttering (7), that Joan is The Real Distinction between Descriptions and Indexicals 53 mayor of Barcelona. Similarly, an utterance of (8) is an assertion of a proposition to which the noun phrase ‘mayor of Barcelona’ does not contribute; in addition, a second proposition is expressed in uttering it, that that is mayor of Barcelona. Dever explains in this way two sorts of intuitive data. First, if the main clause contains operators like ‘might’ in (7) and (8), the material inside appositives or in the noun phrases complementing complex demonstratives is not understood as falling inside their scope. Thus, compare (7) and (8) to (9), one of whose readings is that there is a possible situation such that the mayor of Barcelona there, whoever he is, studies Dennett’s philosophy:

(9) The mayor of Barcelona might study Dennett’s philosophy

The second set of data explained by the proposal are intuitions suggesting that only cross-clause (“E-type”) anaphora can exist between expressions in the main clause and expressions in appositives or the complements of complex demonstratives; compare (10), which has a reading in which ‘them’ is a variable bound by ‘most’, with (11) and (12), which only admit E-type anaphoric readings of ‘them’.3

(10) Most UB logicians danced with students who take courses with them

(11) Most UB logicians danced with this student who takes courses with them

(12) Most UB logicians danced with Marta, who takes courses with them

Bach (1999) also advances a multi-propositional account of conventional implicatures.4 In my own proposal indexical and demonstratives occurring in “ground- level” assertoric utterances (for present purposes, not in direct or indirect contexts or negative existentials) like (1)-(3) contribute their referents to the asserted content. However, they also contribute to the expression of a different proposition; this additional proposition is not asserted, but presupposed. For (1)-(3), the additional proposition is what Stalnaker [Stalnaker (1978)] calls the diagonal proposition expressed by (4)-(6), on the assumption that the indexicals occurring in them are the very same expressions occurring in (1)-(3).5 (The reader can take these expressions to be Kaplan’s [Kaplan (1989a)] “types-in- context”, or more simply tokens, assuming the Reichenbachian view of context-dependence that I prefer.) That is to say, it is the proposition that assigns to any possible world “considered as actual” containing that very utterance in the same language the truth-value that it would have if evaluated relative to 54 Manuel García-Carpintero that world.6 These propositions are necessary; this is how the present account explains the apparent tautological character of (4)-(6). They are necessary in that they would express truths relative to the possible worlds including a context where they are aptly uttered, to the extent that in that context they are still governed by the constitutive principles of English. Similarly, if evaluated relative to the relevant diagonal propositions, the arguments (1)-(3) are nec- essarily truth-preserving, i.e., they are truth-preserving relative to all worlds including contexts like those just described; this is how the account explains their intuitive validity. The conception of presupposition that the account assumes is not the semantic conception, according to which a presupposition is a requirement for an utterance to express a truth-valued proposition. It is rather the pragmatic conception developed by Stalnaker [Stalnaker (1973, 1974, 1978)]. Well-behaved contexts in which utterances are made include a class of mutually known propositions constituting a common ground. Some expressions are properly used only in contexts where the common ground includes specific propositions; these are their presuppositions. Failure in this respect can give rise to different sorts of infelicities, including that the main intended speech act lacks a truth-evaluable propositional content; but there are other possibilities, outstanding among them Lewis’ [Lewis (1979)] accommoda- tion, i.e., that the proposition is added to the common ground so that the presupposition is then satisfied. In some cases, including those that we are here interested in, presuppositions are conventionally associated with expressions; these are not semantic presuppositions in the usual sense, but it is a semantic matter (in the previously explicated sense) to account for them.7 It should be noted that contexts might change even mid-sentence. Thus, the context for interpreting the first occurrence of “this tree” in an utterance of ‘this tree is this tree’ differs from the one for the second, which is why its truth is not linguistically required and why any such utterance can be infor- mative to competent speakers. This applies to the common ground; it of course changes from utterance to utterance (an undisputed assertion adds a new proposition to the common ground), but it can also change mid-sentence, from one expression to another inside the utterance. I hope this is enough for the reader to grasp the form of the asymmetry thesis that I want to defend. Against proposals like those by Larson & Segal [Larson & Segal (1996)], indexicals and demonstratives are not Millian expressions; they semantically describe their referents as having special identifying properties, as (4)-(6) witness (as I said, in my view the same applies to proper names). But these expressions are directly referential: their contribution to the main proposition signified in ground-level utterances including them is their referent. What about definite descriptions? Here I do not have anything original to say. As argued by Neale [Neale (1990)], I think that the best treatment should be given in the framework of the theory of generalized The Real Distinction between Descriptions and Indexicals 55 quantifiers; this is in agreement with Russell’s core view, minus its most glaring shortcomings. I disagree with Neale in thinking –– with, e.g., Heim and Kratzer [Heim and Kratzer (1998)] –– that quantifiers, including definite descriptions, also have semantically associated presuppositions. For present purposes, and assuming we can put the asymmetry thesis thus:

(AT-p) Indexicals occurring in assertoric utterances of ground-level sentences contribute their referents to the asserted proposition; a definite description the NP contributes a relation obtaining between the ex- tension of NP, if it meets the presupposed assumption that its cardinality is 1, and any other class including it.

This is only a preliminary version of the asymmetry thesis; the part concerning descriptions will be modified presently, to take into account the fact that the use of quantificational expressions involves reference to a domain of discourse. (AT-p) accounts for the fact that indexicals, unlike definite descriptions, are de iure rigid designators: if they designate an object o, they designate it with respect to every possible world in which o exists, and do not designate anything else with respect to any possible world; thus, compare intuitions concerning possible-world truth conditions:

(13) That first Spaniard to win the Tour de France was born in Cuenca

(14) The first Spaniard to win the Tour de France was born in Cuenca

According to the proposal, (13) is felicitously uttered only in contexts in which it is known of the referent of the complex demonstrative that, in every world considered as actual, it was a Spaniard first to win the Tour de France; thus, in the actual context the referent is Bahamontes. Hence, (13) asserts something false but contingently so, true of worlds w such that, for instance, Bahamontes’ parents move from Toledo to Cuenca just before he is born; fully independently of the place where the first Spaniard to win in w the Tour de France, if any, was born in w, and fully independently also of whether in w the very same English token of (13) is uttered, and where is it that the referent of the token of the complex demonstrative in that utterance was born in w. Matters are different, however, with (14). It also asserts something false, and contingently so; but the worlds of which it is true include those where Bahamontes does not win the 1959 Tour de France, nor any other, and it is instead Ocaña, in actuality the second Spaniard (born in 56 Manuel García-Carpintero

Cuenca) to win the tour de France, who comes first. This is the intuitively correct theoretical diagnosis of (13) and (14) truth-conditions. As Donnellan showed, ordinary practices include many uses of definite descriptions that intuitively do not differ from referential expressions. Imagine, for instance, that (14) is uttered in the course of a (not very well-informed) bi- ography concerning Bahamontes, with the speaker obviously taking for granted that he shares with his audience the knowledge that Bahamontes was the first Spaniard to win the tour de France, and using the description merely as a sty- listic alternative to the already much-used name. It is clear in that case that the description behaves with respect to the content that the speaker obviously wants to assert as a rigid designator. Grice [Grice (1969)] and Kripke [Kripke (1977)] famously argued, however, that these are cases of speaker’s reference, not semantic reference, which do not therefore contradict AT. Donnellan cases, according to Grice and Kripke, can be explained pragmatically: a speaker can use a definite description non-literally in a directly referential way; even in that case, though, the use of the description in the required “secondary” literal act of meaning is quantificational.8 Interesting cases for Grice’s theory of implicatures are those like the temporal sense of conjunction, the exclusive sense of disjunction, the existen- tially committed sense of the universal quantifier, and so on. There typically are two semantic hypotheses to account for them: a claim of semantic ambiguity, and the rival Gricean view, which contends that the alleged semantic ambiguity does not exist and explains the evidence to the contrary as a case of generalized implicature. Without denying that there are clear examples of generalized non-literal meanings (Grice’s example shows this: expressions- type of the form X is meeting a woman tonight, implying that the person to be met is someone other than X’s wife, his mother or his sister), it has frequently been pointed out that Grice’s criteria do not provide a decision for these cases. The reason is that the applicability of Grice’s tests (derivability from the maxims, cancellability and non-detachability are the main ones) is compatible in these cases with the truth of both premises. Even in cases of ambiguity, the intended sense is “derived” with the help of conversational maxims; the intended proposition is “cancellable” (thanks to the fact that the uttered sentence has at least two possible meanings), etc. [Sadock (1978) and Walker (1976)]. It cannot be denied that there are indeterminate cases. The decisive criterion to be used, however (of which derivability, cancellability and non- detachability are only symptoms), as made clear by both Grice and Kripke, is a methodological one; it is, in effect, as indicated by Grice, an application of Occam’s razor, a “Modified Occam’s Razor”: “Senses are not to be multi- plied beyond necessity” [Grice (1978), p. 47]; “Do not posit an ambiguity unless you are really forced to, unless there are really compelling theoretical or intuitive grounds to suppose that an ambiguity really is present” [Kripke The Real Distinction between Descriptions and Indexicals 57

(1977), p. 259]. The specifically linguistic rationale behind this criterion is, I believe, the following. The point under discussion (whether or not a semantic ambiguity exists) concerns the existence or non-existence of a given semantic convention. Now, conventions differ from other regularities in human behaviour (like digestion, death, or wearing clothes in cold weather), among other things, in one crucial factor: namely, that they are sustained by a particular mechanism, involving complex attitudes (beliefs, expectations and intentions) of a group of people.9 To show that, while one of the alleged senses is undoubtedly conventional, the existence of the other can be explained without positing a specific regularity sustained in that complex way is thus a good theoretical indication favouring the more economical hypothesis. Kripke [Kripke (1977), pp. 257-62)] advances an insightful way of showing this: to imagine a language in which an expression is stipulated to have by convention only the first sense, while convincing oneself that the other sense would still be acquired by the expression by the operation of Gricean conversational maxims –– a mechanism which is over and above the conventions constituting any particular language. Grice [Grice (1969)] and Kripke [Kripke (1977)] provide a pragmatic explanation for Donnellan cases that fit well the previous example. Here is a brief summary. There are circumstances such that the speaker and his audience mutually know both (i) that the expression of a singular content (as opposed to the general one expressed by the uttered descriptive sentence taken literally) is conversationally required, while (ii) that the speaker is not in a position to use a conventional device for doing so: perhaps the intended object is not available to be referred to by means of an indexical, or there is no mutually known proper name. When, in circumstances such as those, the singular proposition concerning the object that would have been referred to by means of an indexical or proper name that it is the one and only ϕ belongs to the conversational background of mutually accepted facts, an assertion including ‘the ϕ’ will be conversationally taken to assert a singular content, stronger than the quantificational content of the literal act, through the mechanism described by Grice [Grice (1975)]. According to it, a literal act of meaning (of only secondary interest for the speaker) has also taken place, to explain how the primary, non-literal act of meaning occurs; the contribution of the description to this secondary, literal act of meaning is still purely quantificational. Kripke [Kripke (1977)] worried that the pragmatic account could be applied to referential uses involving incomplete definite descriptions, as in (15):10

(15) The cup belongs to Marta

Wettstein [Wettstein (1981)] based on Kripke’s worry an argument against AT. However, Blackburn [Blackburn (1988)] showed that there are 58 Manuel García-Carpintero also incomplete but clearly quantificational uses of descriptions –– and Neale [Neale (1990)] did the same regarding quantifiers in general. As Blackburn points out, it is important to recall here that the descriptive material may include straightforward singular terms; regarding the contribution of ‘the Earth’, ‘the satellite of the Earth is small’ is a singular state of affairs, even if, with respect to the contribution made by the descriptive phrase including the name, the asserted content is purely quantificational. Thus, prima facie at least, incomplete definite descriptions do not tell against AT. The best treatment of incomplete descriptions can be established by a form of argument originally due to Partee [Partee (1989)] that Stanley [Stanley (2000)] and Stanley and Szabó [Stanley and Szabó (2000)] have used to argue that quantificational expressions contribute to logical forms hidden indexicals signifying a class, a domain or universe of discourse. Their argument relies on the existence of sentences with readings that are best explained on the assumption that, as befalls to overt indexicals, the assumed hidden indexicals become bound variables. (16) illustrates this:

(16) In every room in John’s house, exactly three bottles are green

Suppose that we assume that there is a hidden indexical u associated with the SD ‘exactly three bottles’, which denotes a domain of discourse in simpler sentences like ‘exactly three bottles are green’ (so that they can be literally true, in spite of the fact that the whole universe contains many more green bottles than three). We can then contend that the hidden indexical be- comes in utterances like (15) a bound variable, obtaining a possibly different value for every value assigned to the variable bound by ‘every’. On the assumption that definite descriptions are quantifiers, (9) also serves to make the point; in the reading such that the description is understood as inside the scope of ‘might’, the domain with respect to which the description is interpret varies with every value assigned to the variable bound by ‘might’. According to Stanley and Szabó [Stanley and Szabó (2000)], this hidden indexical/variable is in fact a nominal restriction, which in logical form cohabits the terminal node occupied by the noun whose projection restricts the quantifier. In my own view, it rather cohabits in LF the DP-projection of the quantifier. Two considerations support this. First, it makes good psychological sense to associate hidden variables with concrete expressions; for logical forms are at least tacitly known by speakers, and we should be able to explain where their constituents come from in ways compatible with this. (Partee’s original examples involve expressions like ‘local’, ‘neighbour’, ‘en- emy’, and so on.) Second, quantifiers occur without being restricted by a noun phrase, while a specific domain of discourse is nonetheless understood. Stanley and Szabó [Stanley and Szabó (2000)] argue that a proposal like mine vio- lates compositionality, because the semantic value of a SD will not just depend The Real Distinction between Descriptions and Indexicals 59 on the semantic values of the nodes that it dominates, but I think that this criticism presupposes an unnecessarily restrictive compositionality principle.11 In sum, the version of the asymmetry thesis that I will defend is this:

(AT) Indexicals occurring in assertoric utterances of ground-level sentences contribute their referents to the asserted proposition; a definite description the NP contributes a relation obtaining between the exten- sion of NP, if it meets the presupposed assumption that its cardinality relative to the tacitly signified domain of discourse is 1, and any other class including it.

II. SCHIFFER’S ARGUMENT

Schiffer [Schiffer (1992)] criticizes this proposal concerning incomplete descriptions on the basis of what he calls “the meaning-intention problem,” the problem that one may reasonably doubt that speakers “mean what the hidden-indexical theory requires them to mean” when they use descriptions [Schiffer (1992), p. 512]. Schiffer contemplates the rejoinder that speakers do have the relevant intentions, albeit somehow “tacitly,” and it is only tacit intentions that semantic theories require them to have: “The best reply for the HI theory is to contend that [speakers] mean it tacitly, where tacit belief is unavailable to consciousness and is not rationally incompatible with thoughts available to consciousness” [Schiffer (1992), p. 512]. He questions this proposal in the following terms. A quantificational theory of incomplete descriptions requires that completing descriptive material be somehow added to the proffered description, to obtain the expression of a property which can be sensibly considered to uniquely specify an object. However, in most contexts there are several candidates that could do the job (“the cup closer to us”, “the red cup”, the “funny-looking cup”); and the tacit intentions of the speaker hardly help to select one of them. The theory, however, claims that it is the descriptive material in question which goes into the individuation of the asserted proposition; it is left, therefore, with the difficulty that no determinate proposition seems to be expressed: “[no candidate] is sufficiently salient to enable you, my audience, to identify it as the one I meant, and this notwithstanding the fact that we understand all the concepts involved and everything has been raised for us to the level of conscious awareness” [Schiffer (1994), p. 287]. Later, however, Schiffer [Schiffer (1994), p. 321] rejected this argument. On behalf of AT, he offers a line inspired by Blackburn [Blackburn (1988)]. “Simply revise the hidden-indexical theory of descriptions so that it allows for the speaker to be making an indeterminate statement. […] In the Pergola example, you did not definitely mean any general proposition in ut- 60 Manuel García-Carpintero tering “The guy is drunk”, but you sort-of-meant, or vaguely meant, several general propositions, one for each definite description that could be used to sharpen what you vaguely meant. And your indeterminate statement might reasonably be held to be true just in case it is true under every admissible sharpening of what you meant, false just in case it is false under every such admissible sharpening, and neither true nor false if it is true under some admissible sharpenings while false under others” [Schiffer (1994)]. The idea is to have recourse to the “supervaluationist” strategy enlisted by Kit Fine [Fine (1975)] to account for the semantics of vagueness. I am not opposed in principle to adopting this strategy. I think, however, that the view outlined in the previous section can help us to account for the facts without positing as much indeterminacy as Schiffer assumes. When a manifestly incomplete description the ϕ is successfully used in an utterance whose logical form has the description dominated by the domain-indexical u, additional token-reflexive descriptive material is implicitly understood, which, together with ϕ, provides the intended description.12 This conventionally understood, implicit descriptive material can be made explicit by means of a description like the ϕ in the most salient domain u when that token of ‘the ϕ’ is produced. Schiffer considers a proposal like this; he rejects it on account of two problems. This is the first: “This suggestion presupposes that the description the most salient F will be the most salient of the candidate definite descriptions, but this is simply false: the fact that the notion of salience occurs in a description does not make that description the most salient description” [Schiffer (1994), p. 115)]. Any plausibility that this objection may have, I think, depends on misinterpreting the proposal. The proposal is not that the additional descriptive material ϕ in the most salient domain u when that token of ‘the ϕ’ is produced is a completion of ϕ that would come easily to ordinary speaker and their audiences. The idea is rather that the proposed characterization provides an accurate abstract theoretical representation of the general strategy that the speakers and their audience rely upon to find the completing material in particular cases; it is, as it were, a higher-level property of the additional descriptive material, a determinable to be determined in each context. This higher-level property guides language users, in particular cases, to find specific determinates, typically by appealing to non-conventional, contextual information. Consider first the case of demonstratives, analogous in respects relevant to the present concern. Demonstratives, as indicated earlier, are indexicals that must be accompanied by a demonstration to make a truth-conditional contribution. Now, demonstrations can obviously be manifested in many different ways. As Evans [Evans (1982)] pointed out, to demonstratively say of someone who just noisily fainted that he was hungry, the use of a pointing gesture is rather superfluous. The resources available for demonstrating cannot in fact be enumerated in advance; the concept of a demonstration is mani- The Real Distinction between Descriptions and Indexicals 61 festly “open-textured”. There exists, however, an accurate characterization of the linguistic rule regarding ‘he’, which could go like this: any instance i of ‘he’ refers to the male demonstrated at the occasion of the production of i. This of course does not mean that “being demonstrated when i occurs” is one more of the criteria which could manifest the relevant directing intentions of the speaker, competing with others when there is some indeterminacy. This theoretical characterization is an acceptable abstract description of the rule tacitly followed by competent users of instances of ‘he’, a determinable to be contextually determined. What has been said here regarding the role of ‘demonstrated’ in the rule for ‘he’ and other demonstratives applies mutatis mu- tandis to the role of ‘salient’ in the characterization of the domain providing completers for successful uses of incomplete descriptions. Think as an example of an utterance of ‘the former Secretary of State writes about the EU’, introducing an article by Kissinger, or one of ‘the Minne- sota butcher did not kill his wife’ as a headline. If one asks the utterer, or the audience, “What ‘former Secretary of State’?” or “What ‘Minnesota butcher’? There are too many!” (or if one consults his untutored intuitions), one would get answers like: “Well, the Secretary of State among authors of articles in this newspaper”, or “the Minnesota butcher among people who has been in the news in the last few days”. The present proposal is that, in articulating answers like these, speakers are tacitly “guided” by the explicitly stated conventional rule; so that they also “know” that it is the secretary of state –– or the Minne- sota butcher –– in the most salient domain in the context of the utterance who are intended. “Being most salient domain in the circumstances” is, as it were, a tacitly understood determinable helping speakers to obtain the relevant determinates in each context; although it is the determinates which would come to mind more easily, and only reflection sophisticatedly theoretical (including the sort of data that Stanley and Szabó [Stanley and Szabó (2000)] consider) could bring out the more abstract description. To claim that the rule we conventionally follow when we use or interpret manifestly incomplete descriptions is the one just given is not to say that, if prompted (say, as part of an informal poll conducted in the classroom), competent speakers (or audiences) would complete without further ado their incomplete descriptions with it. It is to say rather that the rule offers the best theoretical general characterization of our clear intuitions about concrete examples (and whatever other empirical data could be relevant). We are not satisfied with those theoretical claims because they are confirmed by our intuitions (or by an informal poll), but on account of complex theoretical considerations and arguments; among them, comparisons as to how our proposal fares with respect to alternative accounts, and more holistic considerations about the way they interact with other relevant theoretical claims. Schiffer poses a second problem for a quantificational account of incomplete descriptions along the lines of the one just advanced: “The descrip- 62 Manuel García-Carpintero tion the most salient F is itself incomplete: most salient in what respect? I strongly suspect that when this is spelled out the intended description will amount to the circular the F to which I am referring in this utterance of ‘the F’” [Schiffer (1994)]. The first part of this contention has already been granted, in taking ‘saliency’ to be a determinable. However, I do not see any justification for Schiffer’s suspicion in the second half of the quoted material. This can be seen from the fact that the incomplete description can be clearly quantificational, so that it does not even make sense to contemplate an intended “reference” on the part of the speaker. Think of an utterance of ‘the mayor has always been under strong pressure from the land-owners’, in a context where it is perfectly clear that several mayors at different times in the past, and not just the current one, are encompassed in the assertion; the intended logical form could be captured by something like for any relevant time-interval in the past t, the mayor at t was under strong pressure from the land-owners. Here, the implicit descriptive material provided by the present proposal indeed abstractly characterizes the more specific description intended in the context (say, the mayor of Barcelona at t, for specific assign- ments of relevant time-intervals to the variable ‘t’, where Barcelona is the city in the salient domain of discourse). Schiffer’s suspicion of a vicious cir- cularity is out of place; for the speaker is not in any way attempting to refer here to any particular mayor: this is a clear-cut quantificational use of the description. Let us then assume that incomplete definite descriptions can be understood according to the preceding proposal, and let us call the claim that there are no semantic referential uses of definite descriptions the Grice-Kripke view.13 According to the Grice-Kripke view, a literal speaker of ‘the guy is drunk’ in a Donnellan case is using “the guy” in a quantificational way; the asserted proposition, so far as the contribution of ‘the guy’ to it is concerned, is general. Because the assumption is that the use is “referential”, the non- literal, primary act of meaning does involve a singular proposition; but this is irrelevant for understanding the secondary, literal act of meaning. On the other hand, according to the directly referential view of indexicals, a literal speaker of ‘he is drunk’ asserts a singular proposition. Schiffer’s (1995) is not directly concerned with the asymmetry thesis, but with the semantics of attitude ascription and compositional semantics. Here I will be discussing what he calls “first dilemma”, which substantiates his challenge to the asymmetry thesis: he concludes from it that there are directly referential uses of descriptions. The dilemma purports to establish that we cannot coherently combine the Grice-Kripke view with the claim that indexicals are directly referential expressions. Schiffer presents it in three premises. We have just decided to accept the first, which states what we are calling the Grice-Kripke view:

The Real Distinction between Descriptions and Indexicals 63

(i) If the hidden indexical theory of descriptions is correct, then in uttering a sentence containing a referentially used incomplete definite description a literal speaker performs an indeterminate (or determinate) description-theoretic act of meaning [Schiffer (1995), p. 118].

The only qualification needed here is that, in the framework we have developed, the description-theoretic act of meaning is not “indeterminate”, except in the sense that the completing descriptive material is theoretically characterized in general terms, with the help of the determinable ‘salient’: for all that has been argued, the “indeterminacy” at stake is not vagueness, but mere non-specificity. Schiffer’s second and third premises are these:

(ii) If in uttering a sentence containing a referentially used incomplete definite description a literal speaker performs a description-theoretic act of meaning, then so does a literal speaker who utters a sentence containing a referentially used indexical [Schiffer (1995)].

(iii) But if the indexical speaker performs a description-theoretic act of meaning, then there is no evident way one can nonarbitrarily choose between the direct-reference theory of indexicals and the hidden-indexical description theory of indexical [Schiffer (1995)].

“Hidden-indexical description theory of indexicals” is Schiffer’s term for any view that denies what I, following Kaplan, defended in the first section, namely, that an utterance u of “he is drunk” of which i is the relevant instance of ‘he’ asserts a singular proposition. Such a “hidden-indexical description theory of indexicals” makes the utterance synonymous with “the male demonstrated when i is produced is drunk”, assuming a quantificational interpretation of the latter sentence. Thus, as indicated earlier, Schiffer’s dilemma comes down to this: it is not coherent to defend at the same time the Grice- Kripke view of descriptions, and a “directly referential” view of indexicals. As against this, I will argue to defend the asymmetry thesis that Schiffer’s argument commits a fallacy of equivocation.14 Schiffer’s key turn of phrase, “performance of a description-theoretic act of meaning by a literal speaker,” is problematically ambiguous; it can be understood in two different ways. (Henceforth, I will refer to the performance of a description-theoretic act of meaning by a literal speaker as “the problematic notion”.) I will argue that we cannot interpret the phrase as it appears in the second and the third premise in a uniform way, while still regarding both premises as acceptable. A sufficient condition for the literal speaker using an incomplete description to perform a descriptive-theoretical act of meaning is that the asserted proposition be the general state of affairs posited by quantificational accounts. This 64 Manuel García-Carpintero is what obtains, and thus guarantees the performance of such a descriptive- theoretic act of meaning, according to the Grice-Kripke view, whenever a description is used, no matter if the description is incomplete or complete, and no matter if a non-literal singular state of affairs is also meant. We obtain the two interpretations of the problematic notion relative to whether or not we take this sufficient condition to be also necessary. Let me explain. I argued in the first section that when an indexical is used, there is also descriptive material that is semantically involved in the act. The claim made by the defender of the directly-referential view of indexicals, however (as I have theoretically characterized his views anyway), is that this descriptive material does not go into the individuation of the asserted proposition, but is merely presupposed material. This may or may not be sufficient to say that the literal speaker of an indexical sentence “performs a description-theoretic act of meaning” too. Schiffer’s paper is not clear about this, of course; the required distinctions have been made with the essential help of the framework I outlined. Let us say that, in the strict sense, for the literal speaker to perform a description-theoretic act of meaning by means of an utterance including a given term (a description or an indexical) is for him to use the term in a quantificational way: as far as the term’s contribution is concerned, the asserted proposition is general. The sufficient condition in the preceding paragraph, that is to say, is also necessary. Now, if we understand the problematic notion according to this strict sense, Schiffer’s third premise is correct, but the second premise is false; or so I will argue. Let us call the other possible interpretation of the phrase the lax sense; in this sense, and relative to the theoretical framework just developed, the literal speaker of ‘He is drunk’ also performs a description-theoretic act of meaning, just because ‘he’ is associated with a presupposed descriptive mode of presentation which allows the object going into the asserted singular proposition to be identified. This interpretation is lax because the reason why the problematic notion is applied to the literal speaker of “the guy is drunk”, and the reason why it is applied to the literal speaker of “he is drunk” are (if AT is true) substantially different. If we take the notion in the lax sense, the second premise is unobjectionable; it is the third that will then be argued to be false, although for the same reasons as the second is false when the strict interpretation is given to the problematic notion. Let me quote in full the crucial passages of Schiffer’s justification for his second premise:

[…] the two cases have exact psychological parity with respect to those psychological facts on which the relevant speaker meaning would have to supervene. If you ask the speaker what she meant in uttering “The guy is drunk”, you will not get a report that favours the description theory: the speaker will almost cer- tainly offer up an object-dependent proposition involving Pergola, the intended The Real Distinction between Descriptions and Indexicals 65

referent of her utterance of “the guy”. If a theorist is to be justified in discover- ing an indeterminate description-theoretic act of meaning in the utterance of “The guy is drunk”, it will have to be on the basis of the fact that the speaker intended it to be mutual knowledge between her and her audience that certain definite descriptions applied to Pergola and that certain of these were essential to the communicative act in that the speaker would not have uttered her sentence if she had thought those descriptions were not mutually known to be instan- tiated. If we are warranted in ascribing an indeterminate description-theoretic act of meaning to the speaker when she utters “The guy is drunk”, then it will have to be solely on the basis of these psychological facts. But these very same facts also obtain when the speaker utters “He is drunk”. For example, in neither case would the speaker have produced her utterance if she had not thought that just one man was staggering up to the podium to give a talk. Moreover, these psychological facts pertaining to contextually-relevant definite descriptions apply to any utterance of an indexical-containing sentence, and I shall assume that this is clear to you on reflection [Schiffer (1995), p. 120].

I will briefly outline my line of reply before developing it in more detail. Suppose that we interpret the problematic notion in the strict sense. Then, the second premise is false, and what is wrong with Schiffer’s argument (from the perspective of someone holding the Grice-Kripke view) is this part of his justification: “If we are warranted in ascribing an indeterminate description-theoretic act of meaning to the speaker when she utters “The guy is drunk”, then it will have to be solely on the basis of these psychological facts.” (My emphasis; disregard the allegedly indeterminate character of the act.) The theorist with whom Schiffer is arguing wants to claim that, even though “The guy is drunk” works non-literally exactly as “He is drunk”, literally they function differently: the first involves a quantificational use of the description, and therefore it signifies a general fact; the second involves a directly referential use of the indexical, and therefore the signified fact is singular. This is so, even though, confusingly, in the indexical case the proposition expressed involves descriptive elements as well; for those elements are, in the indexical case, only part of presupposed propositions. Because of this, the an- tecedent in Schiffer’s second premise is true, while the consequent is false (given that the problematic phrase receives the strict interpretation). Assum- ing the lax interpretation, on the other hand, it is the conditional constituting the third premise that is false, for the same reasons. At bottom, the problem lies with Schiffer’s assumptions regarding “those psychological facts on which the relevant speaker meaning would have to supervene”. The quoted text betrays that Schiffer has a view of them inappro- priately narrow; for he is prepared to count only propositional attitudes fairly accessible to the speakers as such psychological facts. The defender of the asymmetric treatment of descriptions and indexicals along the Grice-Kripke line, however, correctly contends that, over and above them, the cognitive 66 Manuel García-Carpintero facts that have to be posited to give a plausible general account of language mastery have to be taken into account, even if these psychological facts re- main tacit and are only accessible on the basis of complicated theoretical reflection. This is the point I will be developing. Schiffer seems to be partly aware of it when he says, while discussing the third premise, that the theorist who wants to defend that the description in “The guy is drunk” is quantificational, while the indexical in “He is drunk” is directly referential “can try to deny […] that the two competing hypothesis [namely, the direct-reference view of indexicals and the “hidden-indexical description” view of them] do explain the speaker-meaning facts equally well. […] our theorist might argue that indexicals do not have purely attributive uses, but that they ought to if the hidden-indexical hypothesis is correct” [Schiffer (1995), 122-3]. Here Schiffer manifests some awareness of the sort of facts that, over and above those on which he focus, the development of a correct theory of natural language have to take into consideration. Let us turn to our problem with this in mind. It is clear that descriptions have quantificational uses; no amount of Gricean ingenuity guided by the assumption that they have directly referential uses could explain the way the italicized descriptions work in examples like the following (under the relevant interpretation):

(17) This semester, every professor has given his best student the maximum degree

(18) The mayor of Barcelona has always been under strong pressure from the land-owners.15

These are perfectly common examples of a phenomenon that is ubiqui- tous in natural language.16 Nothing similar can be said regarding indexicals. Notice that the point is not one about frequency. The implicatures we are considering, if they are implicatures at all as contended by the Grice-Kripke view, are generalized –– which means that referencial uses occur quite frequently. But there are undisputed generalized implicatures –– as Grice’s ‘a woman’ example shows. On behalf of the “hidden indexical description” theory of indexicals, Schiffer claims that the same applies to indexicals, even if in a restricted form: “[I]ndexicals arguably do have attributive uses. For example, upon en- countering a huge footprint in the sand, you might exclaim, “He must be a giant!”, and arguably what you would mean is that the man whose foot made the print, whoever he is, must be a giant” [Schiffer (1995), p. 123]. Actually, this is not a very good example. The reason is that the modal ‘must’ is epistemic, and it is at least arguable that singular terms in these contexts behave as they do in indirect contexts. Given that it is a part of the partially Fregean The Real Distinction between Descriptions and Indexicals 67 viewpoint on indexicals I hold that the descriptive material associated with indexicals which is merely presupposed in ordinary context can nonetheless become a constitutive part of the asserted state of affairs when indexicals are embedded in indirect contexts, it is only to be expected that an indexical might behave descriptively when it lies inside the scope of an epistemic modal.17 Nevertheless, there are good instances of the phenomenon Schiffer at- tempts to exemplify. Nunberg [Nunberg (1993)] provides interesting cases, like ‘I am traditionally allowed to order whatever I like for my last meal’, said by the condemned prisoner, or ‘Today is always the biggest party night of the year’. Following Kaplan [Kaplan (1989)], let us call those uses of indexicals “monsters”. In my view, however, the two cases are not at all on a par. Quantificational uses of definite descriptions are common-place; examples like the ones I have given in the previous paragraph are paradigmatic cases of literal uses of language, if there are any literal uses at all. Against Nunberg’s claims to the contrary, however, his instances of (pragmatic) monsters –– as Kaplan suggests [Kaplan (1989), fn. 34] –– can be reasonably seen as non-literal uses.18 Schiffer might reply that we have only statistics to support this. But the claims under dispute are not based on statistics; they are claims concerning what is the best theoretical reconstruction of the compositional rules that con- stitute languages. (On the assumption, empirically as well supported as any empirical hypothesis I know of, that there indeed are linguistically constitutive compositional rules.) We can use Kripke’s strategy, to illustrate how a language stipulated to be such that only directly referential uses of indexicals were conventional could occasionally give rise to monsters. But there is no plausibility in the suggestion that quantificational uses of descriptions could be explained similarly, under the assumption that, conventionally, descriptions only have directly referential uses. In so far as language includes recur- sive rules, descriptions can be recursively embedded inside the scope of any number of those operators –– quantifiers, modalities, temporal expressions, and so on ––, and they do not literally refer in those cases. The initial assumption of a Gricean derivation of an alleged non-literal quantificational use of ‘the mayor or Barcelona’ or ‘his best student’ in (17) and (18) above, that, in his secondary literal use, the speaker intended to refer to a particular student or mayor, does not take us anywhere. To emphasize the importance of these examples for our argument, let us consider a different recent argument by Reimer [Reimer (1998)] for the existence of directly referential uses of descriptions. The argument is ultimately based on the sheer frequency of Donnellan cases: according to Reimer, the Grice-Kripke line does not apply to uses occurring quite frequently. She does acknowledge that frequency is no guarantee of literalness, and she mentions, as an example against the opposite suggestion, the fact that sentences of the form Could you do x? are frequently used to mean Do x; she is prepared to 68 Manuel García-Carpintero grant that, this notwithstanding, these sentences literally query about the hearer’s ability to do x [Reimer (1998), p. 95, fn. 17]. What else is then required for literalness, in addition to frequency? I can only find this suggestion in the paper: “for a (simple or complex) expression e to be capable of literally expressing p, it must be the case that p is appropriately constrained by the linguistic meaning(s) of the expression(s) comprising e” [Reimer (1998), p. 95]. But of course, this only states what is at stake; it cannot serve as criterion to answer our question: are Donnellan uses “appropriately constrained” by directly referential linguistic meanings associated with the descriptions, or are they not? Now, Reimer has devised a theory which, if it applied to English, would have the consequence that Donnellan uses are “appropriately constrained” by directly referential linguistic meanings; and (as far as I am able to discern) she appears to think that the sheer existence of such a theory, together with the facts of frequency, does suffice to disregard the Grice-Kripke line of reply, and to sustain her claim [Reimer (1998), pp. 96-9]. But, of course, a semantic theory can be easily designed, according to which ‘could’ has a different linguistic meaning in utterances of the form Could you do x? To conclude the semantic ambiguity of utterances including ‘could’ in Eng- lish only on that basis, plus the facts of frequency, is in fact to obtain that conclusion just from the latter facts. And to reject the conclusion for the case of ‘could’, while supporting it for descriptions on the indicated basis, is un- motivated. Reimer does object to the Grice-Kripke line of reply when frequent uses are at stake that, by appealing to similar considerations, we would wrongly disregard the facts of frequency on the basis of which we consider dead metaphors aspects of literal meaning [Reimer (1998), p. 98]. If the Grice-Kripke considerations suffice to disregard the frequency of Donnellan cases as an indication of literalness –– of the existence of directly referential uses –– there must be something else, not present in the case of dead metaphors. But there is something else; something analogous to what justifies Reimer in taking the line she takes in the case of Could you do x?: when the expressions occur in utterances that differ only structurally, then the otherwise suspect uses do not exist any more. In the case of Could you do x?, it is enough to change the time to the past, or the person to the third. In the case of descriptions, it is enough to consider complex sentences, in which descriptions interrelate with other operators. “The difficulty [for the Russellian] … is that conditions sufficient for positing of a semantically significant referential interpretation of definite descriptions appear to obtain. If the Russellian disagrees with this claim, then he will be obliged to specify what further conditions would be required” [Reimer (1998), p. 96]. The conditions allegedly sufficient are the frequency of Donellan uses, plus the existence of a coherent theory accounting for them as literal. The defender of AT indeed disagrees. A further condition he would The Real Distinction between Descriptions and Indexicals 69 mention is the fact about the structurally related sentences, together with the fact that we have to account compositionally for the semantics of the expressions at stake. These facts do not apply to dead metaphors; they do apply to the ‘could’ case –– and to descriptions. As one more consideration supporting his anti-asymmetry view, Schiffer also says “there is evidently an explanation of why, given that the hidden- indexical hypothesis is correct, purely attributive uses of indexicals are uncom- mon. It is the same explanation of why it is unusual to find purely attributive uses of maximally incomplete descriptions such as “the guy” or “the table”. For such expressions there is generally no way of raising meaning-candidate complete definite descriptions to contextual saliency except in the presence of a contextually-salient reference candidate” [Schiffer (1995), pp. 123-4] But, firstly, it is not at all unusual to find straightforwardly quantificational uses of maximally incomplete descriptions. If one utters (19), say, while discussing the renovation of the furniture in room 212, such a use is manifestly quantificational:

(19) The table has always been metallic, now we want to try wood19

Secondly, if there were very specific indexicals, purely attributive uses should be, according to Schiffer’s explanation, more common for them. There are such indexicals. Schiffer seems to forget that the contrast we discuss does not simply concern pure indexicals, but demonstratives in general; and among demonstratives there are plenty which are very specific: ‘that unpretentious wooden table which belonged to Wittgenstein’, say. As far as I can tell, however, the frequency of purely attributive uses in those cases is not higher. These considerations take us to what I feel to be the ultimate source of the views I am criticizing. Schiffer says the following, which I endorse: “In any act of communication the speaker will expect numerous propositions to get conveyed to her audience, and many of these are such that the speaker would not have spoken as she did if she had thought that those propositions were either false or not going to be conveyed. But this does not show that those propositions are in any relevant sense meant. Meaning requires special audience-directed intentions, and we must not confusedly elevate psychological presuppositions of an act of meaning into further acts of meaning” [Schiffer (1995), p. 125]. But then he goes on to say: “the same holds when, in the same circumstances, the speaker says ‘The guy is drunk’ instead of ‘He is drunk’. For speaker meaning supervenes on the speaker’s propositional attitudes and the relevant propositional attitudes are precisely the same in the two cases” [Schiffer (1995), pp. 125-6]. This is what we have been given no reason to accept. If it sounds plausible, it is because we are misled into a too narrow concentration on the ordinary speaker’s conscious perception of the situation. The difference between the two cases (indexicals and descriptions) may not show up “atomistically” in this way, when we just consider the spe- 70 Manuel García-Carpintero cific meaning-intentions of the speaker relative to those two utterances; it is only apparent when we consider in general the attitudes constituting speakers’ semantic knowledge. In other words, the difference only shows up when we try to elaborate the best theoretical account of the tacit attitudes manifest- ing themselves in the speakers’ general dispositions to use descriptions and indexicals; the former but not the latter typically and quite literally admit of general readings. There is a final point I will make before concluding. Schiffer’s argument is in fact ultimately intended as an argument against compositional semantics. Because of this, I think that when he disregards the kind of psychological facts I have argued to be really relevant for semantics he is begging the main question at stake. For the argument for compositional semantics is crucially based on the assumption that there are further psychological facts to be taken into consideration in the cases under dispute than those that Schiffer contemplates. Our discussion makes clear the extent to which these issues turn indeed on whether or not natural languages have a compositional semantics. In considering just atomistic issues regarding the specific meaning- intentions that speakers have in mind when they utter sentences like ‘he is drunk’ and ‘the guy is drunk’, while disregarding more holistic issues about how these semantic attitudes of speakers relate to the ones they have with respect to other sentences in which expressions of those kinds also occur, Schiffer’s argument against compositional semantics begs the most important question at stake.

Departament de Lògica, Història i Filosofia de la Ciència Universitat de Barcelona Baldiri Reixach, s/n, E-08028 Barcelona E-mail: [email protected]

Notes

* The present paper derives from discussions with Stephen Schiffer about its main subject; I am very grateful to him for his useful commentaries and suggestions. I am also indebted to comments by different members of Logos, University of Barcelona, on a previous version of the paper, including Jose Díez, Dan López de Sa, Josep Macià, Manuel Pérez and David Pineda, and to commentaries and suggestions by Claudia Bianchi on the occasion of a presentation of this material in a Barcelona-Milano meeting. Thanks to Michael Maudsley for his grammatical revisions. This work, as part of the European Science Foundations EUROCORES Programme OMLL, was supported by funds from the Spanish Government’s grant DGI BFF2002-10164 and the EC Sixth Framework Programme under Contract no. ERAS-CT-2003-980409, from DGI HUM2004-05609- The Real Distinction between Descriptions and Indexicals 71

C02-01, DURSI, Generalitat de Catalunya, SGR01, and a Distinció de Recerca de la Generalitat, Investigadors Reconeguts 2002-2008. 1 I have elaborated on this view in García-Carpintero (2001). 2 These points are further developed in García-Carpintero (2004). 3 Assuming Neale’s (1990) view of E-type anaphora, the referent of ‘them’ could be fixed by the descriptions the UB logicians. In the bound-variable reading of ‘them’ in (10), there are situations of which the sentence is true and each dancing logician danced with students who only take courses with that logician. 4 I take my proposal below about conventional implicatures to be very close in general outline to Barker’s (2003) view. This may come as a surprise, given Barker’s criticism of Bach. I think those criticisms are based on a confusion of what I charac- terize as asserted content vis-à-vis what is said, in the proper semantic sense; the latter is multi-propositional in character, and on some respects goes beyond what is asserted, while in some others fall short of it. I hope to elaborate on this elsewhere. 5 García-Carpintero (forthcoming) elaborates on this, providing reasons why, semantically, the presupposition in question is the diagonal proposition. 6 Talk of worlds “considered as actual” vs. “considered as counterfactual” comes from Davies and Humberstone [Davies and Humberstone (1980)], an alternative way to Stalnaker’s (1978) to deal with Kripke’s “reference-fixing” descriptive material associated with referential expressions. 7 Stalnaker agrees: “Among the reasons that a pragmatic presupposition might be required by the use of a sentence, by far the most obvious and compelling reason would be that the semantic rules for the sentence failed to determine a truth value for the sentence in possible worlds in which the required presupposition is false” [Stalnaker (1973), p. 452]. See the first section of von Fintel (2004) for a recent sketch about the nature of presuppositions and their impact on the semantics-pragmatics distinction very much along the lines suggested here. 8 This terminology (to call the literal act of meaning that is also performed in non-literal acts of meaning ‘secondary’, in contrast to the “primary” non-literal act of meaning intended by the speaker) comes from Searle (1979). 9 I rely on Lewis’s classic analysis in Lewis (1983). 10 In view of the fact that descriptions in examples like the one above, “the cup belongs to Marta,” can be unproblematically understood as helping to make a definite claim about a specific cup. Of course, the problem is only posed by successful incomplete descriptions: those used so that speakers’ intuitions do not manifest any failure of reference. 11 See also Williamson (2003), p. 461, fn. 7, for further support. I think that no difficulty results if we interpret the principle of compositionality as a supervenience principle: utterances cannot differ in meaning, unless some of the lexical units or modes of syntactic composition constituting them differ in meaning; this interpretation has been proposed by Szabó (2001). 12 The resulting description, of course, could still be satisfied by more than one entity in the domain; in that case, the presupposition is not satisfied, and no proposition is expressed. As I said earlier, we only have to account for successful incomplete descriptions: those used in cases in which ordinary speakers do not perceive any reference-failure. 13 Kripke’s stated view is just that Donnellan cases do not establish the opposite. 72 Manuel García-Carpintero

14 I will also indicate later that, in so doing, Schiffer is begging the question that is for him ultimately at stake, namely, the compositional nature of natural languages. 15 Read with the logical form indicated above, for any relevant time-interval in the past t, the mayor of Barcelona at t was under strong pressure from the land- owners –– which makes it clear that we intend to generalize over office-holders, as opposed to referring to the office. 16 It was Evans [Evans (1982), ch. 2] who first argued that examples like these provide the best argument in favour of quantificational accounts. See Neale (1990), p. 47, for other examples. 17 Which is to say, I do believe in Kaplan’s (1989, sec. VIII) (semantic) “monsters”; see Schlenker (2003) for compelling arguments. 18 Nunberg (2004) grants this. The main point of the paper gives an interesting twist to our discussion, on which I cannot properly dwell here. Nunberg provides examples suggesting that the descriptive interpretations available for indexicals which his examples show, whether or not is a semantic phenomenon, is not available for referentially used descriptions; and he mounts an argument for the Russellian view of descriptions on this observation. 19 As before, the intended logical form is something like for any relevant time- interval in the past t, the table [at room 212] at t was metallic (which makes it clear that we intend to generalize over tables, not to refer to a filler of a furniture-role in an office).

REFERENCES

BACH, K. (1999), “The Myth of Conventional Implicatures”, Linguistics and Philosophy 229, pp. 327-66. BARKER, S. (2003), “Truth and Conventional Implicature”, Mind 112, pp. 1-33. BLACKBURN, W. (1988), “Wettstein on Definite Descriptions”, Philosophical Studies 53, pp. 263-78. DAVIDSON, D. (1979), “Moods and Performances,” in Margalit, A. (ed.), Meaning and Use, Dordrecht, D. Reidel; also in his Inquiries into Truth and Interpretation, Oxford, Oxford University Press, pp. 109-21. DAVIES, M. AND HUMBERSTONE, L. (1980), “Two Notions of Necessity”, Philosophical Studies 38, pp. 1-30. DEVER, J. (2001), “Complex Demonstratives”, Linguistics and Philosophy 24, pp. 271-330. DONNELLAN, K. (1966), “Reference and Definite Descriptions”, Philosophical Review 75, pp. 281-304. EVANS, G. (1982), The Varieties of Reference, Oxford, Clarendon Press. FINE, K. (1975), “Vagueness”, Synthese 30, pp. 265-300. GARCÍA-CARPINTERO, M. (1998), “Indexicals as Token-Reflexives”, Mind 107, pp. 529-63. — (2000), “A Presuppositional Account of Reference-Fixing”, Journal of Philosophy XCVII (3), pp. 109-47. — (2001), “Gricean Rational Reconstructions and the Semantics/Pragmatics Distinction”, Synthese 128, pp. 93-131. The Real Distinction between Descriptions and Indexicals 73

— (2004), “Assertion and the Semantics of Force-Markers”, in Bianchi, C. (ed.), The Semantics/Pragmatics Distinction, CSLI Lecture Notes, Chicago, The University of Chicago Press, pp. 133-66. — (forthcoming), “Two-dimensionalism: a Neo-Fregean Interpretation”, in García- Carpintero, M. and Macià, J. (forthcoming) (eds.), Two-Dimensional Semantics, Oxford, Oxford University Press. GRICE, H. P. (1969), “Vacuous Names”, in D. Davidson and Hintikka, J. (ed.) Words and Objections, Dordrecht, D. Reidel Publishing Company, pp. 118-45. — (1975), “Logic and Conversation”, in Cole, P. and Morgan, J. (eds.), Syntax and Semantics, vol. 3, New York, Academic Press. Also in Grice, H.P., Studies in The Ways of Words, Cambridge, Mass., Harvard U.P., pp. 22-40, to which I refer. — (1978), “Further Notes on Logic and Conversation”, in Cole, P. (ed.), Syntax and Semantics, vol. 9, New York, Academic Press. Also in Grice, H.P., Studies in The Ways of Words, Cambridge, Mass., Harvard U.P., pp. 41-57, to which I refer. HEIM, I., and KRATZER, A. (1998), Semantics in Generative Grammar, Oxford, Blackwell. KAPLAN, D. (1989), “Demonstratives”, in Almog, J., Perry, J. and Wettstein, H. (eds.), Themes from Kaplan, Oxford, Oxford University Press. KRIPKE, S. (1977), “Speaker’s Reference and Semantic Reference”, in French, P., Uehling, T. and Wettstein, H., Contemporary Perspectives in the Philosophy of Language, pp. 6-27, Minneapolis, University of Minnesota Press. Also in Martinich, A.P. (ed.) (1990), The Philosophy of Language, 2nd. ed., Oxford, Oxford U.P., pp. 248-67, from where I quote. — (1979), “A Puzzle about Belief”, in Meaning and Use, Margalit, A. (ed.), Dordrecht, D. Reidel, pp. 239-83. LARSON, R. and SEGAL, G. (1996), Knowledge of meaning: Semantic value and Logical Form, Cambridge, MIT Press. LEWIS, D. (1979), “Scorekeeping in a Language Game”, Journal of Philosophical Logic (8), pp. 339-59. Also in Lewis, D. (1983), Philosophical Papers, vol. 1, Oxford, Oxford University Press, pp. 233-49. — (1983): “Languages and Language”, in Lewis, D., Philosophical Papers, vol. 1, Oxford: Oxford University Press. NEALE, S. (1990), Descriptions, Cambridge, Mass., MIT Press. NUNBERG, G. (1993), “Indexicals and Deixis”, Linguistics and Philosophy 16, pp. 1-43. — (2004), “Descriptive Indexicals and Indexical Descriptions”, in Reimer, M. and Bezuidenhout, A. (eds), Descriptions and Beyond, Oxford, Oxford U.P. PARTEE, B. (1989), “Binding Implicit Variables in Quantified Contexts”, Papers from the Chicago Linguistics Society 25, pp. 342-65. REIMER, M. (1998), “Donnellan’s distinction/Kripke’s test”, Analysis 58, pp. 89-100. SADOCK, J. (1978), “On Testing for Conversational Implicature”, in Cole, P. (ed.), Syntax and Semantics, 9, New York, Academic Press, pp. 281-97. SCHIFFER, S. (1992), “Belief Ascription”, Journal of Philosophy LXXXIX, pp. 499-521. — (1993), “Actual-language Relations”, in Tomberlin, J. (ed.), Philosophical Perspec- tives 7: Language and Logic, Atascadero, Cal., Ridgeview Pub. Co., pp. 231-58. — (1994), “A Paradox of Meaning”, Nous XXVIII, pp. 279-324. — (1995), “Descriptions, Indexicals, and Belief Reports: Some Dilemmas (But Not the Ones You Expect)”, Mind 104, pp. 107-31. 74 Manuel García-Carpintero

SCHLENKER, P. (2003), “A Plea for Monsters”, Linguistics and Philosophy, 26, pp. 29-120. SEARLE, J. (1979), “Referential and Attributive”, Monist 62, pp. 190-208. STALNAKER, R. (1973), “Presuppositions”, Journal of Philosophical Logic 2, pp. 447-457. — (1974), “Pragmatic Presuppositions”, in Munitz, K. M. and Unger, P. K. (eds.), Semantics and Philosophy, New York University Press, New York. Also in Davis, S. (ed.), Pragmatics, pp. 471-81, to which I refer. — (1978), “Assertion”, in Cole, P. (ed.) Syntax and Semantics 9, New York, Academic Press, pp. 315-32. — (1997), “Reference and Necessity”, in A Companion to the Philosophy of Language, eds. Wright, C. and Hale, B., Oxford, Blackwell, pp. 534-54. STANLEY, J. (2000), “Context and Logical Form”, Linguistics and Philosophy 23, pp. 391-434. STANLEY, J. and SZABÓ, Z. G. (2000), “On Quantifier Domain Restriction”, Mind and Language 15, pp. 219-61. VON FINTEL, K. (2004), “Would You Believe It? The King of France is Back! (Presuppositions and Truth-Value Intuitions)”, in Reimer, M. and Bezuidenhout, A. (eds), Descriptions and Beyond, Oxford, Oxford U.P., pp. 269-96. WALKER, R. (1975), “Conversational Implicature”, in Blackburn, S. (ed.): Meaning, Reference and Necessity, Cambridge, Cambridge U. P., pp. 133-81. WETTSTEIN, H. (1981), “Demonstrative Reference and Definite Descriptions”, Philosophical Studies 40, pp. 241-57. WILLIAMSON, T. (2003), “Everything”, in Hawthorne, J. (ed.), Philosophical Perspec- tives, 17, Language and Philosophical Linguistics, pp. 415-65.